In the technical field of human-computer interaction, with the development of interactive games, virtual reality and head-mounted display technologies, operations that are performable in free movement without usage of hands will be applied more widely. Therefore, identification of a human body posture is the most convenient, flexible, accurate and reliable human-computer interaction method, except for that achieved with a keyboard, a mouse and a touch screen. Existing human body posture identification methods are mainly implemented by using a third-party sensing device to sense a motion of the head at a certain distance. However, due to a limited detection range of the sensing device, the existing human body posture identification methods do not have a good mobility; moreover, the sensing device is sensitive to environmental impacts and susceptible to disturbances, thus causing an error in sensed data.